An Integer Program based Appliance Identification and Load Disaggregation Method

Obtaining information on electricity consumption and operating habits of appliances on the user side of the grid is crucial for optimizing energy rationing and promoting smart grid development. This paper focuses on the household scenario and proposes a two-step non-intrusive appliance load monitoring (NIALM) model, targeted on low-frequency (around or less than 1 Hz) load data using event detection and integer program. The first step is to detect appliance switchover and to reshape the load curve into a step-like curve. Based on the reshaped curve, an unsupervised appliance identification model is proposed, combining integer program and event-based method. In the second step, integer program is used for load disaggregation. Our method is tested on a private dataset and the Reference Energy Disaggregation Dataset (REDD). Experiments demonstrate that our method can successfully identify appliances from load profiles and disaggregate them from the load curve. The performance of our method is computationally efficient and promising for practical use.